Project Summary The human brain tracks multiplexed signals during social interactions. The breakdown of any of these computations could lead to social deficits observed in many psychiatric disorders. While social neuroscience has been growing rapidly in recent years, the complexity of human social interactions has not been well quantified with computational models. Importantly, previous social neuroscience research generally assumes that the structure of social environments are stochastic and social agents act in a reactive way, leaving at least two knowledge gaps in the literature: 1) the proactive nature of social agents and 2) the dynamic and multidimensional feature of social space. The overarching aim of this project is to develop novel computational models and paradigms to capture social controllability and social navigation in ?unselected? human participants (laboratory study n=100, mobile app n=10,000), which can ultimately be used to capture social failures across disorders. In Aim 1, we will develop a novel generative model and paradigm for social controllability, based on a rich literature on model-based decision-making and our previous work on social learning. Key subject-level parameters include: simulated controllability (delta), future thinking weight (i.e. weight put on planning future interactions), and learning rate (epsilon). In Aim 2, we will delineate navigational computations of dynamic social relationships using a novel social interaction game in which participants interact and develop relationships with virtual characters, we will devise novel measures that track the trajectories of social relationships and geometrically quantify the overall structure of individuals? two-dimensional social space framed by power and affiliation. In Aim 3, we will use machine learning to 1) deep phenotype participants along the dimensions of social controllability and navigation and 2) predict clinical and subclinical symptoms among a large sample of ?unfiltered? volunteers. Upon successful completion of these aims, this proof-of-concept project will provide important validation for new computational frameworks for social controllability and social navigation, potentially breaking new grounds for computational psychiatry research of social dysfunction. The resulting paradigms, models, and findings will be critical for a wide range of clinical disorders including psychotic, mood, and personality disorders. Furthermore, the proposed paradigms can be back-translatable to animal models, in relation to the social defeat model of depression and other animal models of social behaviors. Thus, the proposed computational framework could have far-reaching influences that would exceed the specific focus on social control and social space navigation, advancing the possibility to advance mechanistic understanding of and develop individualized diagnosis and treatments across multiple psychiatric disorders.